Information Technology Reference
In-Depth Information
the base cases based on the analysis on the goals features (Keane 1988).
Keane names the goal features as structural index in analogical retrieval.
However, can we ignore those base cases without any goal relevance? In
fact, goal features is just an important additional restraint. If we emphasize
goal features too much, we might lose the potential useful cases with similar
sub structures by the filter of goal features, which is unwise.
5.7.4
Individual similarity
Another emphasis in our model is the restraints posed by the classification
information about individuals. Informally speaking, if two individuals have
some similarities, then they belong to the same classification. In the concept
clustering, we cluster the concept set in terms of the relevance or closeness
among them. Relevance refers to the average of similar degrees between the
features of the concepts. But here, we still cluster electric wires and rope into
the same class, because they both can be used to tie up things.
Sometimes, some individuals in a case play the primary role in the
problem solving. In those situations, this kind of cases should be used as the
primary retrieval information to search the case base tentatively. After the
finish of tentative retrieval, we should pay more attention to the cases within
the same class as the target case or having whole-part relations with the target
case. The analogy among individuals helps us to identify them. The partial
solutions to the case can help to find the whole solution.
5.7.5
Similarity assessment
(1) The similarity among numeric features.
Sim V V
(
,
)
=
1
−
d V V
(
,
)
=
1
−
d
i
j
i
j
ij
1
1
Sim
(
V
,
V
)
=
=
or
i
j
1
+
d
(
V
,
V
)
1
+
d
i
j
ij
d
=
V
−
V
ij
i
j
